Claude Opus 5 (xHigh): benchmark results

Provider: Anthropic. Released 2026-07-24. Access: API.

Unified ELO 1791 ± 1, rank #10 of 1761 rated models, from 14 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
ProgramBench4.5Resolved (%)100
ProgramBench Almost37Almost (%)100
Creative Writing (Lechmazur)4.1Mean Score98.9
DuelLab Overall76DuelLab Score98.2
Design Arena (Agents - Mobile Apps)1298Elo97.7
MCP Atlas85.8Pass Rate (self-reported)97
LLM2014 Logic 2026-0964.42Median Score95.5
SEAL - MCP Atlas85.8Score93.3
LLM2014 Logic 2026-0864.74Median Score93.2
LLM2014 Logic 2026-0771.88Median Score93
Epoch AI - Critpt27.71Score92.6
CursorBench 3.169.3Score (%)90.4

Interactive version: theaggregate.ai/model?slug=claude-opus-5-xhigh · How It Works · Data refreshed daily, snapshot 2026-09-05.